mirror of
https://github.com/openssl/openssl.git
synced 2024-12-27 06:21:43 +08:00
1064616012
Add Neon path for RSA on armv8, this optimisation targets to A72
and N1 that are ones of important cores of infrastructure. Other
platforms are not impacted.
A72
old new improved
rsa 512 sign 9828.6 9738.7 -1%
rsa 512 verify 121497.2 122367.7 1%
rsa 1024 sign 1818 1816.9 0%
rsa 1024 verify 37175.6 37161.3 0%
rsa 2048 sign 267.3 267.4 0%
rsa 2048 verify 10127.6 10119.6 0%
rsa 3072 sign 86.8 87 0%
rsa 3072 verify 4604.2 4956.2 8%
rsa 4096 sign 38.3 38.5 1%
rsa 4096 verify 2619.8 2972.1 13%
rsa 7680 sign 5 7 40%
rsa 7680 verify 756 929.4 23%
rsa 15360 sign 0.8 1 25%
rsa 15360 verify 190.4 246 29%
N1
old new improved
rsa 512 sign 12599.2 12596.7 0%
rsa 512 verify 148636.1 148656.2 0%
rsa 1024 sign 2150.6 2148.9 0%
rsa 1024 verify 42353.5 42265.2 0%
rsa 2048 sign 305.5 305.3 0%
rsa 2048 verify 11209.7 11205.2 0%
rsa 3072 sign 97.8 98.2 0%
rsa 3072 verify 5061.3 5990.7 18%
rsa 4096 sign 42.8 43 0%
rsa 4096 verify 2867.6 3509.8 22%
rsa 7680 sign 5.5 8.4 53%
rsa 7680 verify 823.5 1058.3 29%
rsa 15360 sign 0.9 1.1 22%
rsa 15360 verify 207 273.9 32%
CustomizedGitHooks: yes
Change-Id: I01c732cc429d793c4eb5ffd27ccd30ff9cebf8af
Jira: SECLIB-540
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14761)
181 lines
6.0 KiB
Plaintext
181 lines
6.0 KiB
Plaintext
LIBS=../../libcrypto
|
|
|
|
$BNASM=bn_asm.c
|
|
IF[{- !$disabled{asm} -}]
|
|
# Define source files and macros per asm architecture
|
|
# Known macros are:
|
|
#
|
|
# OPENSSL_BN_ASM_PART_WORDS For any collection with /-586/ file names
|
|
# OPENSSL_BN_ASM_MONT For any collection with /-mont/ file names
|
|
# OPENSSL_BN_ASM_MONT5 For any collection with /-mont5/ file names
|
|
# OPENSSL_BN_ASM_GF2m For any collection with /-gf2m/ file names
|
|
# OPENSSL_IA32_SSE2 For any collection with /86/ file names
|
|
# when sse2 is enabled
|
|
# BN_DIV3W For any collection with /-div3w/ file names
|
|
#
|
|
# All variables are named in such a way that they can be "indexed" with
|
|
# $target{asm_arch}
|
|
|
|
$BNASM_x86=bn-586.s co-586.s x86-mont.s x86-gf2m.s
|
|
# bn-586 is the only one implementing bn_*_part_words
|
|
# => OPENSSL_BN_ASM_PART_WORDS
|
|
$BNDEF_x86=OPENSSL_BN_ASM_PART_WORDS OPENSSL_BN_ASM_MONT OPENSSL_BN_ASM_GF2m
|
|
$BNDEF_x86_sse2=OPENSSL_IA32_SSE2
|
|
|
|
$BNASM_x86_64=\
|
|
x86_64-mont.s x86_64-mont5.s x86_64-gf2m.s rsaz_exp.c rsaz-x86_64.s \
|
|
rsaz-avx2.s rsaz_exp_x2.c rsaz-avx512.s
|
|
IF[{- $config{target} !~ /^VC/ -}]
|
|
$BNASM_x86_64=asm/x86_64-gcc.c $BNASM_x86_64
|
|
ELSE
|
|
$BNASM_x86_64=bn_asm.c $BNASM_x86_64
|
|
ENDIF
|
|
$BNDEF_x86_64=OPENSSL_BN_ASM_MONT OPENSSL_BN_ASM_MONT5 OPENSSL_BN_ASM_GF2m
|
|
$BNDEF_x86_64_sse2=OPENSSL_IA32_SSE2
|
|
|
|
IF[{- $config{target} !~ /^VC/ -}]
|
|
$BNASM_ia64=bn-ia64.s ia64-mont.s
|
|
ELSE
|
|
$BNASM_ia64=bn_asm.c ia64-mont.s
|
|
ENDIF
|
|
|
|
$BNASM_sparcv9=asm/sparcv8plus.S sparcv9-mont.S sparcv9a-mont.S vis3-mont.S \
|
|
sparct4-mont.S
|
|
$BNDEF_sparcv9=OPENSSL_BN_ASM_MONT
|
|
$BNASM_sparcv9_ec2m=sparcv9-gf2m.S
|
|
$BNDEF_sparcv9_ec2m=OPENSSL_BN_ASM_GF2m
|
|
|
|
$BNASM_sparcv8=asm/sparcv8.S
|
|
|
|
$BNASM_alpha=bn_asm.c alpha-mont.S
|
|
$BNDEF_alpha=OPENSSL_BN_ASM_MONT
|
|
|
|
$BNASM_mips32=bn-mips.S mips-mont.S
|
|
$BNDEF_mips32=OPENSSL_BN_ASM_MONT
|
|
$BNASM_mips64=$BNASM_mips32
|
|
$BNDEF_mips64=$BNDEF_mips32
|
|
|
|
IF[{- ($target{perlasm_scheme} // '') eq '31' -}]
|
|
$BNASM_s390x=bn_asm.c s390x-mont.S
|
|
ELSE
|
|
$BNASM_s390x=asm/s390x.S s390x-mont.S
|
|
ENDIF
|
|
$BNDEF_s390x=OPENSSL_BN_ASM_MONT
|
|
$BNASM_s390x_ec2m=s390x-gf2m.s
|
|
$BNDEF_s390x_ec2m=OPENSSL_BN_ASM_GF2m
|
|
|
|
$BNASM_armv4=bn_asm.c armv4-mont.S
|
|
$BNDEF_armv4=OPENSSL_BN_ASM_MONT
|
|
$BNASM_armv4_ec2m=armv4-gf2m.S
|
|
$BNDEF_armv4_ec2m=OPENSSL_BN_ASM_GF2m
|
|
|
|
$BNASM_aarch64=bn_asm.c armv8-mont.S
|
|
$BNDEF_aarch64=OPENSSL_BN_ASM_MONT
|
|
|
|
$BNASM_parisc11=bn_asm.c parisc-mont.s
|
|
$BNDEF_parisc11=OPENSSL_BN_ASM_MONT
|
|
$BNASM_parisc20_64=$BNASM_parisc11
|
|
$BNDEF_parisc20_64=$BNDEF_parisc11
|
|
|
|
$BNASM_ppc32=bn-ppc.s ppc-mont.s
|
|
$BNDEF_ppc32=OPENSSL_BN_ASM_MONT
|
|
$BNASM_ppc64=$BNASM_ppc32 ppc64-mont-fixed.s
|
|
$BNDEF_ppc64=$BNDEF_ppc32
|
|
|
|
$BNASM_c64xplus=asm/bn-c64xplus.asm
|
|
$BNASM_c64xplus_ec2m=c64xplus-gf2m.s
|
|
$BNDEF_c64xplus_ec2m=OPENSSL_BN_ASM_GF2m
|
|
|
|
# Now that we have defined all the arch specific variables, use the
|
|
# appropriate ones, and define the appropriate macros
|
|
IF[$BNASM_{- $target{asm_arch} -}]
|
|
$BNASM=$BNASM_{- $target{asm_arch} -}
|
|
$BNDEF=$BNDEF_{- $target{asm_arch} -}
|
|
IF[{- !$disabled{ec2m} -}]
|
|
$BNASM=$BNASM $BNASM_{- $target{asm_arch} -}_ec2m
|
|
$BNDEF=$BNDEF $BNDEF_{- $target{asm_arch} -}_ec2m
|
|
ENDIF
|
|
IF[{- !$disabled{sse2} -}]
|
|
$BNDEF=$BNDEF $BNDEF_{- $target{asm_arch} -}_sse2
|
|
ENDIF
|
|
ENDIF
|
|
ENDIF
|
|
|
|
$COMMON=bn_add.c bn_div.c bn_exp.c bn_lib.c bn_ctx.c bn_mul.c \
|
|
bn_mod.c bn_conv.c bn_rand.c bn_shift.c bn_word.c bn_blind.c \
|
|
bn_kron.c bn_sqrt.c bn_gcd.c bn_prime.c bn_sqr.c \
|
|
bn_recp.c bn_mont.c bn_mpi.c bn_exp2.c bn_gf2m.c bn_nist.c \
|
|
bn_intern.c bn_dh.c bn_rsa_fips186_4.c bn_const.c
|
|
SOURCE[../../libcrypto]=$COMMON $BNASM bn_print.c bn_err.c bn_srp.c
|
|
DEFINE[../../libcrypto]=$BNDEF
|
|
IF[{- !$disabled{'deprecated-3.0'} -}]
|
|
SOURCE[../../libcrypto]=bn_depr.c bn_x931p.c
|
|
ENDIF
|
|
SOURCE[../../providers/libfips.a]=$COMMON $BNASM
|
|
DEFINE[../../providers/libfips.a]=$BNDEF
|
|
# Because some CPUID implementations use some BN assembler (!!!), we
|
|
# must include assembler code into the legacy provider under the same
|
|
# conditions as CPUID code is included. See ../build.info
|
|
SOURCE[../../providers/liblegacy.a]=$BNASM
|
|
DEFINE[../../providers/liblegacy.a]=$BNDEF
|
|
# Implementations are now spread across several libraries, so the defines
|
|
# need to be applied to all affected libraries and modules.
|
|
DEFINE[../../providers/libcommon.a]=$BNDEF
|
|
|
|
INCLUDE[bn_exp.o]=..
|
|
|
|
GENERATE[bn-586.s]=asm/bn-586.pl
|
|
DEPEND[bn-586.s]=../perlasm/x86asm.pl
|
|
GENERATE[co-586.s]=asm/co-586.pl
|
|
DEPEND[co-586.s]=../perlasm/x86asm.pl
|
|
GENERATE[x86-mont.s]=asm/x86-mont.pl
|
|
DEPEND[x86-mont.s]=../perlasm/x86asm.pl
|
|
GENERATE[x86-gf2m.s]=asm/x86-gf2m.pl
|
|
DEPEND[x86-gf2m.s]=../perlasm/x86asm.pl
|
|
|
|
GENERATE[sparcv9a-mont.S]=asm/sparcv9a-mont.pl
|
|
INCLUDE[sparcv9a-mont.o]=..
|
|
GENERATE[sparcv9-mont.S]=asm/sparcv9-mont.pl
|
|
INCLUDE[sparcv9-mont.o]=..
|
|
GENERATE[vis3-mont.S]=asm/vis3-mont.pl
|
|
INCLUDE[vis3-mont.o]=..
|
|
GENERATE[sparct4-mont.S]=asm/sparct4-mont.pl
|
|
INCLUDE[sparct4-mont.o]=..
|
|
GENERATE[sparcv9-gf2m.S]=asm/sparcv9-gf2m.pl
|
|
INCLUDE[sparcv9-gf2m.o]=..
|
|
|
|
GENERATE[bn-mips.S]=asm/mips.pl
|
|
INCLUDE[bn-mips.o]=..
|
|
GENERATE[mips-mont.S]=asm/mips-mont.pl
|
|
INCLUDE[mips-mont.o]=..
|
|
|
|
GENERATE[s390x-mont.S]=asm/s390x-mont.pl
|
|
GENERATE[s390x-gf2m.s]=asm/s390x-gf2m.pl
|
|
|
|
GENERATE[x86_64-mont.s]=asm/x86_64-mont.pl
|
|
GENERATE[x86_64-mont5.s]=asm/x86_64-mont5.pl
|
|
GENERATE[x86_64-gf2m.s]=asm/x86_64-gf2m.pl
|
|
GENERATE[rsaz-x86_64.s]=asm/rsaz-x86_64.pl
|
|
GENERATE[rsaz-avx2.s]=asm/rsaz-avx2.pl
|
|
GENERATE[rsaz-avx512.s]=asm/rsaz-avx512.pl
|
|
|
|
GENERATE[bn-ia64.s]=asm/ia64.S
|
|
GENERATE[ia64-mont.s]=asm/ia64-mont.pl
|
|
|
|
GENERATE[parisc-mont.s]=asm/parisc-mont.pl
|
|
|
|
# ppc - AIX, Linux, MacOS X...
|
|
GENERATE[bn-ppc.s]=asm/ppc.pl
|
|
GENERATE[ppc-mont.s]=asm/ppc-mont.pl
|
|
GENERATE[ppc64-mont.s]=asm/ppc64-mont.pl
|
|
GENERATE[ppc64-mont-fixed.s]=asm/ppc64-mont-fixed.pl
|
|
|
|
GENERATE[alpha-mont.S]=asm/alpha-mont.pl
|
|
|
|
GENERATE[armv4-mont.S]=asm/armv4-mont.pl
|
|
INCLUDE[armv4-mont.o]=..
|
|
GENERATE[armv4-gf2m.S]=asm/armv4-gf2m.pl
|
|
INCLUDE[armv4-gf2m.o]=..
|
|
GENERATE[armv8-mont.S]=asm/armv8-mont.pl
|
|
INCLUDE[armv8-mont.o]=..
|